WHAT LIES BEHIND THE DEBATE: DIFFERENT VIEWS ABOUT EVIDENCE

On the particularist view, you can have evidence for indi­vidual hypotheses, and you can have evidence for a set of hypotheses, but only if you have evidence for each hypoth­esis in the set.

Both the holist and the particularist, of the sorts I am considering, invoke a concept of explanation in their views of evidence. For both, e is evidence for a set of hypotheses H only if there is some kind of explanatory connection between H and e. Now, one way—perhaps the only way—to generate a holistic doctrine is to understand “explains” in a deduc­tive way, according to which a set of hypotheses H potentially explain e only if H deductively entails e. (On this view, for H to correctly explain e, the hypotheses in H must be true.) And if H potentially explains e, then e is evidence that H. (In Whewell's more sophisticated version, if H deductively explains e, then e is evidence that H, provided that the set H is “consilient” and “coherent.”)

Now, the present holist, who understands “explains” in a deductive way, wants to deal with a set of hypotheses H in which each hypothesis is necessary to derive the results described in e. He doesn't want to say that if H explains e, then so does H conjoined with any arbitrary statement H'. Yet if H entails e, then so does the conjunction H & H'. Indeed, in Hempel's “deductive-nomological” model of explanation, it is required that for a set of hypotheses to potentially explain

some phenomena e, each member of the set must be neces­sary for the deduction. So, on the corresponding deductive- explanatory view of evidence, e is evidence for the set H only if all the hypotheses in H together deductively entail e and no subset in H does.

One problem here is that the deductive account of ex­planation, whether Hempel's or Whewell's more demanding one (requiring “consilience” and “coherence”), is subject to serious counterexamples. These show that the requirements for an explanation proposed by these accounts are neither necessary nor sufficient. (See section 6 below, for examples.)

A potentially more serious problem pertains to the cen­tral idea of defining “evidence” in such a way that if the set of hypotheses H potentially explains some established phe­nomena e, then e is evidence that H. Another way of putting this central idea is that if H, if true, correctly explains e, then e is evidence that H. The problem, which arises whether or not “explains” is understood in a deductive way, is how to re­spond to the “competing explanation” (or “competing hypoth­esis”) objection noted in chapter 3, section 2. Suppose that H, if true, would correctly explain e. In general, the objection goes, there are competing sets of hypotheses which, if they were true, would correctly explain e. Yet it would seem ab­surd to say that e is evidence for each of these competing sets.

Whether the objection is valid or not depends on what else can be said about the original and alternative explanations. When Newton confronts this objection in his discussion of Rule 4, his reply is that if your theory has proper causal and inductive support (that accords with his Rules 1-3), and if the competing theory has no such support, then the fact that there is a competing theory does not, in his terms, “nullify” the causal and inductive support for the original.

Whewell, as Mill notes, omits this causal-inductive re­quirement.

To this we might respond: How does Whewell know this? He doesn't even give an argument! Perhaps Whewell didn't know about, or live long enough to know about, competing versions of the nineteenth-century wave theory, and particularly versions of the ether, that were championed. One invokes a continuous ether, another an ether composed of discrete particles. In an article on the ether, Maxwell says that it might be thought that the fact that the ether is elastic and compressible

is a proof that the medium [the ether] is composed of sep­arate parts having void spaces between them. But [Maxwell continues] there is nothing inconsistent with experience in supposing elasticity or compressibility to be properties of every portion, however small, into which the medium can be conceived to be divided, in which case the medium would be strictly continuous.^[122]

These versions of the wave theory, though conflicting, explain lots of different optical phenomena (including rec­tilinear propagation, reflection, refraction, diffraction, inter­ference, etc.) known during the nineteenth century. Indeed, there are at least three versions of the wave theory: one postulates a continuous ether, one postulates a discontin­uous ether, and one postulates an ether but makes no claim about continuity or discontinuity.

If we seek to employ an explanatory concept of evidence, avoid both the competing-hypothesis objection (in damaging forms), as well as the problems facing a deductive account of explanation, and try to settle the holism-particularism debate, we need to look elsewhere. I shall do so by using one of the explanatory concepts of evidence introduced in chapter 1, viz. (potential) A-evidence,^[123] which I defined as follows:

e is (potential) A-evidence that h if and only if it is more prob­able than not that, given e, there is an explanatory connection between h and e (p(E(h,e)/e) > '/2); e is true; e does not entail h.

This definition avoids the competing hypothesis objection, since if p(E(h,e)/e) > %, then for any competing hypothesis h', p(E(h',e)/e) < Th This definition, I will show, employs a concept of correct explanation that avoids the standard counterexamples to deductive accounts of correct explana­tion. The important question is whether the definition can be used to support holism, particularism, or neither. The an­swer must await section 8.

The definition above appeals to the idea of an explana­tory connection. As indicated in chapter 1, there is such a connection between h and e if and only if h correctly explains why e is true; or e correctly explains why h is true; or some hypothesis correctly explains why both e and h are true. But now we need to dig deeper and provide an account of a “correct explanation” that allows us to address the holism­particularism debate and avoids problems with deductive ac­counts of explanation.¹⁶

5.